Test Impact Analysis for Android & JVM

As your team and product grow, the number of features in development inexorably begins to grow.

All the while, the business wants to get new features to users faster and faster, which means you have to accelerate the regression in some way.

Autotests are great for this. But what to do when you have too many of them? 2, 3, 4, 5, 10 thousand tests, and all these have to be run somewhere just before the regress!

And how do you maintain it all?

This is where impact analysis - a tool that allows us to clearly understand which tests should be run on a pooled road.

Maxim just wrote such a tool - with Jacoco and EmmaRT we instrument the bytecode, map the classes and methods we load at runtime, and then, on the pool-request, parse the git diff and understand what UI tests we need to run.